Which MB for statistical analysis?

[ylimes2]

I know there are a lot of these threads, but I'd really appreciate any thoughts or words of wisdom on the following existential musings on which MacBook to buy. I had mostly decided but since the WWDC keynote I'm having doubts again. I'm leaving the real world this summer to go back to (grad) school, necessitating a laptop purchase, but I'm a couple years out of touch with both the MacBook lineup and student habits.

Most important is which machine, if any, will be competent to run statistical software packages (Stata etc) with large datasets, which I expect to be doing somewhere between occasionally and regularly. The rest of the time I'll just be doing LaTeX and web browsing, which of course anything can do. I also want a machine that will be functional and reliable for some years to come.

Moderately important are weight and, to a lesser extent, battery life. I do like to work in libraries and coffee shops and would prefer to haul the computer around with me most of the time.

I do not care in the slightest about hard drive size, optical drive, graphics, or display.

If a cMBP will be substantially better at running statistical software then that's obviously what I should get, even though it's a bit maddening to buy something right before I expect it to be discontinued. The education pricing on it is great and though it's the heaviest of the 13" Apple notebooks it does still weigh less than either of my previous two laptops. But if the MBP will still struggle (OR not be a huge improvement over the Air's capability, but I am operating under the assumption that the Air will be mostly useless for Stata) and I'm going to end up tied to the university computer lab for running my analyses anyway then I may as well get an Air.

Or should I bite the bullet and shell out a few hundred dollars extra for the rMBP even though I don't actually care about the retina display? Is the SSD really worth the extra cash? It also seems more future-proof than the cMBP. Ultimately I have the money, I'm just not sure the rMBP would be the best use of it.

 

[B...]

Since you don't care about the display, the main benefit of the rMBP would be lost. So I say go for the Air. It is actually a very capable machine which has more raw power now than the rMBP according to Geekbench. The battery life and weight are unbeatable. It is not as "niche" as it was in 2008- it is now a lightweight, relatively powerful notebook.

 

[r6mile]

Get the new 13" Macbook Air. CPU is just as fast, 35% lighter, 40% better graphics, higher screen resolution than the cMBP, 12h battery life, and cheaper.
Stata runs great on an i5.

If you don't care about the display (advantage of the rMBP) or the optical drive (advantage of the cMBP), get the MBA. I'm getting it as soon as Back to School starts.

 

[leman]

13" MBA or 15" rMBP

 

[ylimes2]

Hm, this is not so much what I expected. Thus far I've been considering the Air the least likely option. If I were to go that route, is the upgrade to 1.7/i7 worth it over the 1.3/i5?

----------

13" MBA or 15" rMBP

Care to elaborate on these apparently disparate recommendations? Is this "give up one of your criteria" advice or what?

 

[Stetrain]

I think the main question is whether your application would benefit from a quad core CPU.

If so then you really should go with a 15" MBP, retina or not. The 13" MBPs are not a very big performance increase from the Air, but the 15" models are a bit step up because of the quad core CPU.

SSD is definitely worth it, but you can add your own to the non-retina model.

The retina model basically gives you the same power and the SSD in a thinner lighter package if that matters to you.

 

[leman]

Care to elaborate on these apparently disparate recommendations? Is this "give up one of your criteria" advice or what?

I think the main question is whether your application would benefit from a quad core CPU.

There is your answer

Basically, the question is what kind of statistical analysis are you going to do? Are you doing simulations (e.g. monte carlo methods, resampling or stuff like that) or are you just running Chi2 tests/linear regressions? How big are your datasets? Does your software support multiple cores? And things like that. The difference between the Air CPUs and the one in the MBP 13" is not that big - but the 15" model has quad core CPUs. If you are getting the MBP, I would advice you to take the retina model because of three factors: much faster storage, more real screen estate and increase mobility. The price is not that different if you recall that it comes with a very fast SSD.

As to the i7 vs i5 Ghz CPUs in the Airs - the i7 seems to be around 10-15% faster in synthetic benchmarks so I assume that it will be a similar story for numeric applications.

 

[r6mile]

I am also a postgraduate student, in political science, and at university we run i5s for Stata (with 4GB RAM), which run great even with large datasets.

I honestly don't see the need for an i7, and that seems to be the consensus on the forum unless you're doing lots of rendering, etc. But if you do, the upgrade is not that expensive.

As to whether to get a 13" MBA or a 15" rMBP, I'd go with the MBA for two reasons.
1. Price. The rMBP costs nearly TWICE as much.
2. The performance increase is not something I need, and from your description, you wouldn't either.

I say get the 13 MBA, upgrade to 8GB RAM, and enjoy.

 

[ChrisA]

Most important is which machine, if any, will be competent to run statistical software packages (Stata etc) with large datasets, which I expect to be doing somewhere between occasionally and regularly. The rest of the time I'll just be doing LaTeX and web browsing, which of course anything can do. I also want a machine that will be functional and reliable for some years to come....

Just just did the same thing, back to grad school (after 20+ years) My first question is "how large are large data sets" I've heard some people calling 1,000 rows "large". No those are "trivial" If you are talking millions then the data is still small enough to all fit in RAM. I now reserve the term "large data set" to be the types of data they get from sources like high energy physics where the data sets are measured in terabytes.

So what is "large". The main problem is the small size of the FLASH disk in the Macbokk Air. Will your data fit on a 128GB flash drive? Ithink if the data sets you use are smaller and the files are just a few megabytes yo'd be fine.


The problem with the "Air is the small size of RAM and the small storage. Good enough for Latex, web browsing, writing and reading papers and not taking but look at your data.

Buy the 13" air if you can. the MBP is a bit heavy to be hauling around all day. Heavy enough that I'm temped to leave it home and many ties do. I pretty much depends on the data size.

 

[Laco]

Its also important to consider what version of Stata you have or will buy.

For example, the basic version of Stata is $595 but the version that utilizes 4 cores is $1,695.

http://www.stata.com/order/new/edu/single-user-licenses/#

 

[ylimes2]

I'd chosen not to consider the 15"s for a variety of reasons. In addition to the price jump from 13", the quad core version of Stata is also a great deal more expensive. Since I also lose some portability due to the size then I may as well go hang out in the computer lab at no personal financial cost.

Seems like the difference between the 13" Pros and 13" Air is smaller than I expected so thing to do is save some cash, get an Air (I am planning on 8GB RAM no matter what), and see how it goes with future projects with the understanding that I may not be doing the heaviest analysis on the go or at home.

Thanks all.

 

[Stetrain]

I'd chosen not to consider the 15"s for a variety of reasons. In addition to the price jump from 13", the quad core version of Stata is also a great deal more expensive. Since I also lose some portability due to the size then I may as well go hang out in the computer lab at no personal financial cost.

In that case I would go with the 13" Air. The i7 may be worth the cost, the RAM is definitely worth it.